NVIDIA/DeepStream

NVIDIA DeepStream Monorepo: DeepStream SDK and reference apps for building GPU‑accelerated, real-time video and multi‑sensor analytics pipelines with GStreamer, TensorRT, and vision AI models across edge, on‑prem, and cloud deployments.

What it solves

DeepStream simplifies the development of complex, real-time video and image analytics applications. It eliminates the need to manually manage hardware-accelerated video decoding, encoding, and AI model inference, allowing developers to build high-performance pipelines that process multiple streams simultaneously.

How it works

The toolkit is built on top of GStreamer, a popular multimedia framework. It provides a set of specialized plugins that tap into NVIDIA GPU hardware (dGPU and Jetson) to handle intensive tasks like TensorRT-based AI inference, object tracking, and message-broker integration. Developers can assemble these components into pipelines either through code or by using declarative configuration files and visual tools.

Who it’s for

It is designed for developers and engineers building AI-powered vision applications on NVIDIA hardware, such as smart city monitoring, industrial automation, or security systems. It is particularly useful for those who need to deploy multi-model, multi-stream analytics at scale.

Highlights

  • Hardware Acceleration: Leverages NVIDIA GPUs for efficient video processing and AI model execution.
  • Pipeline Flexibility: Supports complex, multi-stream, and multi-model inference workflows.
  • AI Coding Assistance: Includes a suite of "skills" for AI coding agents to help generate pipelines, profile performance, and onboard vision models.
  • Tooling Ecosystem: Bundles utilities for camera auto-calibration, visual pipeline building, and OpenTelemetry integration for monitoring.
  • Deployment Ready: Provides support for Docker containers and Triton Inference Server for scalable deployment.

Related

  • Project
    iimeta/fastapi

    An enterprise-grade LLM API integration system that provides a unified, OpenAI-compatible interface to access multiple AI model providers.

  • Project
    hardrave/NIGHTRUN

    A local LLM runtime that boots directly from USB or SD card without a conventional operating system, providing a dedicated offline AI interface.

  • Project
    vortico/flama

    Flama is a production framework that packages predictive and generative AI models into portable files and serves them as high-performance APIs with built-in support for agentic tools and standard LLM protocols.

  • Project
    AXERA-TECH/ax-samples

    A collection of reference C++ implementations for deploying deep learning models on AXERA-TECH AI SoC hardware, providing developers with the tools to evaluate and run vision algorithms on edge devices.